OpenAI的菲吉·西莫谈广告为何将登陆ChatGPT

看起来谷歌为了做到这一点不得不尝试了几次，但最终他们做到了。

It seems like Google has had to take a few swings to get this right, but they finally are.

是的，这也很有趣。

Yeah, it's interesting too.

我认为苹果公司的人并不像这样通过软件进行屏幕上的产品叙事。

I think the folks at Apple, they don't really do on screen product storytelling with software in this way.

但他们对AI的尝试，我的意思是，他们在这方面已经有一年半了，比如贝拉·拉姆齐的那个，或者那位忘记送礼物的母亲的案例。

But the tries that they had at AI with AI, I mean, gosh, they're they're a year and a half old at this point, the ones with Bella Ramsey or otherwise of like the mother who forgets a gift.

于是她制作了一个AI记忆视频，呈现得像经过大量努力和真实的情感，这很有用，但并不算酷或富有情感。

And so she makes an AI memory movie, and presents it as a lot of hard work and authentic, like, that's useful, but not really cool or emotional.

而谷歌又一次精准地把握住了，我不知道他们是否还是和同一家广告公司合作，毕竟已经过了这么多年。

Whereas Google just kind of nails it again, I don't know if they're working with the same agency or what after all these years.

我认识一些参与过这类工作的人员，但到目前为止，这已经形成了一种固定模式。

I know some of the folks who've kind of been involved in that type of work here or there, but it's really it's a template at this point.

你只需要想出一个真挚的故事，然后让它自然呈现即可。

And all you have to do is just come up with the heartfelt story and and let it be.

埃利斯，在我们开始与斐济的访谈之前，我有个特别酷的经历，一直想跟你聊聊。

Ellis, quickly before we get to this interview with Fiji, I had a really cool experience that I've been dying to talk with you about.

是吗？

Is that so?

是的。

Yes.

所以上周我在上一期节目中稍微提了一下。

So last week I teased this on the last episode.

我上周在旧金山参加了一个高度保密的简报会，见到了乔尼·艾维和全新的电动法拉利。

I was in San Francisco doing a highly secretive briefing with Johnny Ive and the new electric Ferrari.

我一整天都和乔尼、马克·纽森，还有Love From的其他几位创始人，以及几乎整个法拉利领导团队在一起。

So I spent the day with Johnny, Mark Newsome, his co founder at Love a bunch of other Love From people, pretty much the entire Ferrari leadership team.

他们向我们展示了全新电动法拉利Luche的内饰。

And they showed us the interior of the all new electric Ferrari Luche as it's called.

我通常并不参加汽车试驾活动。

And I do not normally do car junkets.

这实际上是我第一次。

This is actually my first one.

我不报道汽车，但你知道，乔尼的参与让这一切对我而言非常吸引人。

I don't cover cars, but you know, the Johnny of this all made it very intriguing to me.

自从他离开苹果后，我偶尔见过乔尼，但这次是我和他相处时间最长的一次。

And I've gotten to see Johnny a little bit since he left Apple, but this was by far the most time I've spent with him.

我将在采访和消息源中逐步透露这些内容。

And I'm gonna be teasing out kind of all this in the interview and sources.

在本集发布时，这些内容就会公开。

It'll be out by the time this episode drops.

但说实话，这真是一次非常棒的经历。

But yeah, man, it was a very cool experience.

你看过内饰的视频了吗？

Have you gotten to see the video of the interior yet?

我看过了。

I did.

我看了。

I did.

这是他们全新的设计语言。

It's a whole new design language for them.

对我来说，最大的疑问是，尽管大家对回归实体按钮议论纷纷，但这些按钮仍然被设计成屏幕的一部分，我觉得这在某种程度上是有道理的。

I think the biggest question for me was for all the hullabaloo about bringing back buttons, they're still kind of presented as part of the screen, which I think on the one hand makes sense.

你知道，你面前有一个巨大的仪表盘，一部分是屏幕，一部分是按钮。

You know, you kind of have this big dashboard, part of it is a screen, part of it is buttons.

我喜欢他们仍然保留了速度表之类的东西。

I like that they still kept, you know, things like the speedometer.

但我觉得它明显少了点肌肉感，这可能是有原因的。

But it distinctly feels a bit less muscle y, I think, for a reason.

我的意思是，汽车有趣的地方之一，就是那些仪表、旋钮和各种小装置。

I mean, that's one of the fun things with cars, I think, is all the dials and knobs and this and that.

当你在屏幕上看到性能模式，它看起来就像你Mac上的一个小组件时，我就在想，这会不会本来可以设计成另一个仪表盘、旋钮，或者某种你可以通过旋转来调节的东西？你懂我的意思吧？

And when you look at it on that screen and the performance mode looks like a widget on your Mac, I feel like I wonder if that could have been better as as another dial or or a knob or something that you could rev, you know what I'm saying, in a circular motion.

我的意思是，还没达到迷你库珀这些年来所做到的程度。

I mean, not as big as the things Mini Cooper has done over the years.

但确实，我对他们试图在全屏幕和全物理按键之间架起桥梁的方式很感兴趣。

But, yeah, I was interested in kind of the way that they tried to bridge the gap between all screen and all buttons.

是的。

Yeah.

你觉得它还不够模拟，这很有趣，因为相比市场上任何一款豪华电动车，或者大多数新款豪华燃油车，它无疑更接近模拟体验。

It's interesting that you feel like it's still not analog enough because it's definitely a way more analog car than any luxury electric vehicle on the market or most new luxury, even gas vehicles on the market.

所以整个内饰都是铝和玻璃材质。

So the entire interior is aluminum and glass.

约翰尼说，这是他见过的，或者他认识的任何人做过的最精细的CNC加工。

And Johnny was saying it's the most detailed CNC work that he's ever seen or that anyone he's ever worked with has ever done.

这显然意味着他认为这款车的CNC工艺比苹果设备还要好。

So that's obviously saying he thinks this is better CNC work than an Apple device.

考虑到它的售价，本就该如此。

And it should be for what it's going to cost.

我敢肯定这东西的价格会高达五十万美元。

I'm sure this thing's going to cost like half $1,000,000.

他们还没公布价格。

They haven't revealed the price yet.

他们只是给我们展示了内饰。

They were just showing us the interior.

我甚至还没看到这辆车的外观。

I actually didn't even get to see the exterior of the car yet.

他们真的把这辆车像苹果发布会那样拆解了，进去后看到一堆零件摆在苹果风格的桌子上，还有不同的颜色。

They really, they deconstructed this like an Apple event where you go in and there's a bunch of parts laid out on Apple like tables and the different colors.

所有部件都悬浮在房间中央，拆解开来，你可以绕着它走动，仔细观察。

You've got everything like floating in the middle of the room, deconstructed and where you can walk around it and look at it.

天啊，真是太美了。

And man, it's beautiful.

这确实是一个完全不同的方向。

It is definitely a different direction.

我明白你的意思，也许在某些方面它可以更偏向模拟风格，但说实话，我对所有部件的手感印象深刻。

And I hear you that maybe there are areas where it could be even more analog, but I was frankly just blown away with how tactile everything was.

我的意思是，你能明显感受到他们在每一个按钮、旋钮和开关上投入的用心。

I mean, you can really tell the level of thought that they put into every button click, every knob, every switch.

没错，这确实是一种全新的车型。

And yeah, definitely a new kind of car.

是的，我得说明清楚，我知道在播客里我得保持批评态度。

Yeah, I think a lot of it to be clear, I know I have to be critic on a podcast.

很多地方看起来简直美得惊人。

A lot of it looks absolutely stunning.

我的意思是，那个启动按钮，中央控制台上的那个大法拉利按钮，你可以按下去。

I mean, the launch button, the big Ferrari button you can press, like, in the center console.

我超爱方向盘后方的仪表盘。

I love the instrument cluster behind the steering wheel.

那里有个类似航空风格的G力显示装置，简直酷到爆。

There's, like, a fucking aeronautical style g forces thing, like, just absolutely badass.

但你知道这让我想起了什么吗？就是DNA公司的一位设计师，我相信他现在还在那里。

But you know what this reminds me of, which is in love from DNA, one of their designers who I believe is still there.

迈克·马蒂斯，他还在吗？

Mike Mattis, is he still there?

你有没有机会和参与这个项目的人聊过？

Did you get to talk to worked on this.

因为好吧。

Because okay.

看吧？

See?

这让我想起了Nest的圆形界面和这里的一些设计。

It reminded me of Nest, the circular interfaces and some of the stuff here.

我觉得这看起来就像是迈克的DNA风格。

And I'm like, this looks like Mike's DNA.

是的。

Yep.

很高兴你提到这一点。

So glad you brought that up.

曾经有一次，我们和约翰尼以及马克·纽索姆讨论过温控问题，因为温控没有自动设置。

There was a moment where we were talking with Johnny and Mark Newsome about the temperature control because there's no auto setting for the temperature control.

显然，他们曾就是否直接设定温度而没有自动按钮展开过激烈讨论。

And apparently that was a big debate they had of you just set the temperature and that there's no auto button.

约翰尼谈到温度信息的呈现方式时说，是的，我们让迈克负责了这个部分。

Johnny was talking about the way that the temperature information moves and he was like, yeah, we had Mike work on this.

所以，你知道，只有像DNA这样的公司才能做到这一点——让Nest的原始设计师来负责法拉利的温控系统。

So it's, you know, that's like only the thing that's something like love from could do is you have like the original designer of Nest work on the Ferrari's temperature control.

我是迈克的超级粉丝，他做的所有事情我都非常欣赏。

I'm a huge fan of Mike and everything he's done.

对于不了解的人，他设计了大量最初的iPhone应用程序。

For those who don't know, he did a lot of the original iPhone applications.

他还设计了Nest。

He did Nest.

他在其间还创办过其他初创公司。

He's done some other startups in between.

他还做了Facebook Paper，这是我最爱的应用之一。

He also did Facebook paper, one of my favorite apps ever.

至于这些，我觉得Nest在很多方面都做得不够好。

And yeah, I felt like with these I mean, Nest is just just so goaded for so many reasons in terms of what it didn't do.

还有屏幕和可旋转旋钮的结合，仍然保留了恒温器原有的熟悉感和肌肉记忆。

And the mixture of, you know, screens and and turnable rings that still retains the thermostats familiar, you know, muscle memory to it.

是的。

Yeah.

所以看到法拉利的设计语言和他们的设计融合在一起，非常有趣。

So this was super interesting to see that kind of hybrid between the Ferrari design language and theirs.

对。

Yeah.

关键是，这个按键由会变色的墨水制成，当你把它放进车里时会发光。

I mean, the key is made of ink that changes color and lights up when you put it in the vehicle.

你提到的仪表盘旋钮。

The dials you were talking about the instrument cluster.

所以它们是物理旋钮，但其实是数字屏幕。

So they're physical dials, but they're digital screens.

这就像一组四屏的数字屏幕，其间整合了物理仪表。

So that's like a four panel series of digital screens with physical instruments integrated throughout.

他们保留了视差效果。

They kept the parallax effect.

当你从侧面看时，就能看到视差。

So when you look to the side, you see the parallax.

天哪。

Oh man.

对我来说，苹果汽车的议题一直笼罩着整个简报，因为他曾在苹果负责汽车项目。

You know, the Apple Car for me kind of hung over this whole briefing because he obviously worked on the car when he was at Apple.

这个项目持续了大约十年。

It was a project that was going on for like a decade.

苹果从未确认过它的存在。

Apple's never confirmed its existence.

他们据说几年前一度取消了这个项目。

They, you know, reportedly killed it at one point a few years ago.

所以我认为我和在场的其他几位更关注科技的记者都在寻找一些线索，看看这是否是本应属于苹果汽车的精神继承者。

And so I think me and a few of the other tech, more tech minded reporters who were there were kind of looking for strains of, is this the spiritual successor to what would have been the Apple Car?

约翰尼基本上否定了这一点，但还是很难不在这辆车的内饰中看到苹果的影子和那种苹果风格。

And Johnny basically said no, but it's still hard to not see that apple, you know, eye and that apple taste in the products, in the interior.

我的意思是，我们永远无从得知那辆车可能长什么样，但我觉得通过CarPlay和这款车，我们大概已经能很好地推测出驾驶员控制界面原本会是什么样子。

I mean, we're never gonna know about what the car might look like, but I think between CarPlay and this, we probably have a damn good sense of what the actual driver controls may have looked like.

显然，法拉利和苹果原本的优先级会有所不同。

Obviously, the prioritization is going to be a little bit different for Ferrari versus what would have been Apple.

但如果你看看中控面板上那些用于调节温度等功能的小开关，它们确实有实体触感，而在其上方，似乎有一块隐藏在面板后的动态小屏幕，会显示相关信息。

But I mean, if you look at the little switches that they have on the control panel, you know, for the temperature and this and that, there is a tactile piece, but then right above it, there is a little screen seemingly behind that panel that is dynamic and shows you.

所以你确实能感受到实体按键与屏幕动态功能的完美结合。

So you do get like a nice kind of combination of the tactile and the power of screens to be dynamic.

我觉得，你可能会看到类似的东西。

I do think, you know, you would have seen probably something similar.

我的意思是，从约翰尼对这个的评论来看，他们不会把整个系统做成全触屏，正是因为约翰尼所说的，而且他说得对：使用触屏时，你会把视线从路上移开。

I mean, you see with Johnny's quotes about this, like, they wouldn't have made the whole thing a touchscreen because of exactly what Johnny said, and he's right, is that when you're using a touchscreen, you're taking your eyes off the road.

我的意思是，显然我们也可以争论，未来你可能根本不需要盯着路。

And I mean, obviously, you know, we could argue about, well, in the future you're not gonna have to look at the road.

是的。

Yeah.

但不管怎样，

But anyway,

和约翰尼·伊夫的这次体验非常棒。

very cool experience with Johnny Ive.

好了，我想我们终于可以进入对OpenAI应用部门负责人Fiji Simo的采访了。

And with that, I guess let's finally kick it to our interview with OpenAI's CEO of applications, Fiji Simo.

Fiji，非常感谢你邀请我们来到你美丽的家中。

Fiji, thank you so much for having us here at your lovely home.

非常感谢您邀请我参加这个播客。

Thank you so much for having me on this podcast.

能来这里很好。

It's good to be here.

有太多话题可以聊了。

There's so much to talk about.

这周，我觉得尤其是星期五，就在超级碗之前。

This week, I feel like in particular, we're talking Friday right before the Super Bowl.

发生了很多事。

There was a lot.

这周我参加了三次与OpenAI团队的媒体电话会议。

I was on, like, three press calls with the OpenAI team this week.

你们正在发布很多产品。

You guys are shipping a lot.

我想先聊聊广告。

I wanna start with the ads though.

所以Anthropic发布了它的超级碗广告，而萨姆已经做出了回应。

So Anthropic released its Super Bowl ads, and Sam has responded.

你看过这些广告了。

You've seen them.

对吧？

Right?

是的。

Yes.

好吧。

I Okay.

你看过这些广告了。

You've seen them.

我们都看过了。

We've all seen them.

萨姆已经做出了回应。

Sam has responded.

凯特也做出了回应，你的首席营销官。

Kate has responded, your CMO.

但我很想听听你看到这些广告时的感受和想法。

But I'd love to know what you saw what you thought when you when you saw ads.

是的。

Yeah.

好吧，这些广告挺搞笑的。

Well, look, the ads are funny.

就是那些广告。

So those ads.

同时，你知道，我最强烈感受到的是，我们现在正处在一个非常特别的时刻，可以提醒每个人：你就是一个创作者，有了AI，你可以创造任何东西。

At the same time, you know, the thing that I felt most acutely is like, we're at this very interesting moment in time where we can really remind everyone that they're a creator and with AI, you can build anything.

这实际上正是整个行业的核心故事。

Like that's really the story of the entire industry.

而这正是我们超级碗广告想要讲述的故事。

And that's the story of our Super Bowl ad.

我发现，如果能让每个人意识到AI是为了赋予你们超能力，那么AI行业中的每个人都能把蛋糕做得更大。

And I find it that, you know, everyone in the AI industry could make the pie bigger for everyone, if we can make everyone realize that AI is here to give you the superpowers.

我认为，比起仅仅针对竞争对手，这是一个强大得多的信息，当然他们也在这么做，而我们也会这么做。

And I think that's such a much more powerful message than just hitting at a competitor, you know, but they do them, and we're gonna do you do us.

这引发了人们的讨论。

It gets people talking.

确实，这引发了人们的讨论。

It gets people talking for sure.

作为一个准营销理论家，我觉得在超级碗广告中，真正能让人记住的，可能是人们会完全忘记广告本身传达的信息，因为大多数人甚至还没在任何这些服务中体验过。

As a wannabe marketing theorist, I feel like the thing that actually is gonna stick out more to people in the Super Bowl, I think they're gonna forget about the ads message completely because most people haven't even experienced it yet inside of any of these services.

萨姆提到的一个关键点是，德克萨斯州的CHECHBT用户数量比Claude全球的用户还多。

The zinger that Sam had was there were more CHECHBT users in Texas than Claude has globally.

那句话是

That was

是的

Yeah.

但我认为人们会记住的，有趣的是文化层面的那些精湛的演员表演，他们演绎出一种非常典型的、我们常说的谄媚、自信、快乐、机械化的形象。

But I think what people are gonna remember, which is interesting culturally, is just the impeccable actor performances that portrayed this very, like, you know, we say sycophantic, confident, happy, robotic.

当然，你们比任何人都更清楚，这些东西最终会被逐步优化和调整，人们可以自行选择语调。

Obviously, you guys know better than anybody else that that stuff is gonna be fine tuned out and worked through over time, and people can choose the tone of voice.

但我觉得，正是这一点会让人们竖起耳朵、发笑。

But that is what I think is gonna make people perk their ears up and laugh.

而且，如果非要说的话，我认为这甚至可能在科技圈之外盖过广告本身的信息。

And, yeah, if anything, I I think that might actually overshadow the ads message outside of tech.

是的。

Yeah.

至于广告信息本身，我认为这将会引发一场非常有趣的讨论，对吧？

And And, on the ads message itself, I think that's gonna be a very interesting debate, right?

比如，AI究竟是为所有人服务，还是只服务于那些付得起钱的人？

On like, is AI truly for everyone or only for the people who can pay for it?

我认为，我想要生活在一个我们赋予人们选择权的世界里。

And I think I want to live in a world where we give people the choice.

你可以选择付费，也可以选择不付费而看广告。

Like you have the choice to pay if you can, you have the choice to see ads if you don't want to pay.

但至关重要的是，我们要找到一种商业模式，能够将最顶尖的智能普及给每个人。

But it's really important that we find the business model that allows us distribute the best levels of intelligence to everyone.

我认为，如果我们不引入广告，反而把广告描绘成邪恶的东西，我们在人工智能的普及和赋能每个人的愿景上，最终会陷入一个更糟糕的境地。

And I think if we don't introduce ads, and we end up like, portraying ads as like this evil thing, we're going to end up in a much, much worse place in terms of diffusion and like the promise of AI, which is empowering everyone.

所以，我认为这实际上错失了一个机会，没能把行业引导到它应有的位置。

So, I think it's a bit of a missed opportunity to actually set the industry in the place that it needs to be set.

至于语气，是的，我认为这只是一个特定时刻的表现。

And as for the tone, yes, I think it's a moment in time.

这挺有趣的。

It's funny.

我认为我们还在不断调整策略和人格设定，以便每个人都能找到自己喜爱的个性。

And I think we're continuing to iterate on the strategy personality so that everyone finds the personality they want.

但这也正是关键所在，对吧？

But that's also the thing, right?

每个人都想要不同的东西。

Like everyone wants something different.

当我与我们的用户交谈时，有些用户非常喜欢那种非常活泼、能让他们振奋起来的性格。

When I talk to our users, there are some users who love the like super cheerful, like, you know, personalities that pumps them up.

而另一些用户则说：不，不，不，我就想要简洁的一句话。

And then you have some users who are like, no, no, no, I want the one liner.

而关键在于，要为每个人提供工具，让他们能根据自己的喜好进行个性化设置。

And like, the art is gonna be like giving everyone the tools to personalize for what they want.

作为一位元老级的Facebook员工，我想你会理解这一点。

As a OG Facebooker, I think you'll appreciate this.

这周让我大量思考了Facebook与苹果的ATT追踪时代，当时两家公司之间的相互攻击，使用的论点和现在非常相似。

This week has got me thinking a lot about the Facebook versus Apple ATT tracking era where there were the barbs were flying in kind of a similar way between two companies, and they were attacking each other's business models with kind of similar arguments.

现在看到这种情况也发生在AI领域，很有意思。

And it's interesting to see now that that's kind of come to AI.

作为一位在科技行业深耕多年的人，当你看到这种情况再次发生时，你觉得这说明了AI目前处于什么阶段？

And I wonder like, as someone who's been in tech for as long as you have, when you see that happening again, what does that say about where AI is right now?

我的意思是，这正是你所想象的，那就是我们现在正处于扩散和规模最为重要的阶段。

I mean, I think that is exactly what you imagine, which is we're at the point where diffusion and scale is the most important thing.

事实证明，整个行业一再得出结论：广告是一种非常不错的商业模式，能让产品保持免费。

And it turns out we have concluded as an industry over and over again, ads is a pretty good business model to keep products free.

它为科技行业和所有用户创造了巨大的价值。

And it has created an immense amount of value in tech, And for all users.

因此，你知道，对于那些用户基数很小的其他公司来说，可能不需要考虑这一点。

And so, you know, we're at the point where maybe other companies that have very small user bases don't need to think about that.

但对我们而言，当我们触达了8亿用户时，你就得想，我要尽可能多地为每个人提供智能服务。

But for us, when we reached 800,000,000 people, and you're like, I want to give as much intelligence as possible to everyone.

当然，广告也是一个需要考虑的因素。

Of course, ads is a consideration.

不过，我认为真正重要的是要意识到，广告本身也已经进化了，对吧？

Now, the thing I think is really important is to realize that ads has also evolved, right?

我们不会像过去其他公司那样做广告，因为我们拥有一个根本不同的媒介。

Like we are not going to do ads the way other companies in the past have done ads, because we have a fundamentally different medium.

因此，我们在推出广告时确立了非常坚定的原则，比如广告不会影响IGBT给你提供的回复，因为我们认为广告完全独立、并确保用户对这些回复的信任至关重要。

So, we came out when we announced ads with super strong principles, like ads are not gonna influence the response that IGBT gives you because we think full independence from ads and trust that like the trust that people have in these responses is absolutely critical.

所以我们不会照搬过去那种模式。

So we're not gonna like copy paste models of the past.

我们当然不会复制某些广告平台曾经做过的那些糟糕做法。

We're not gonna, you know, obviously reproduce all of the bads that some ads platform have done.

我们希望打造一种完全不同的广告平台，它非常适合当前这个时刻、这款产品，因为AI是不同的。

We want to create a very different kind of ad platform that is very appropriate for this moment, this product, and AI is different.

我认为，长期在Snapchat从事广告业务甚至参与广告业务运营，让我感受到创始人——包括我今天合作的那些人——往往倾向于做出宏大、高尚的承诺。

I think covering ads and even working within an ads business at Snapchat for such a long time, I feel like there's often a tendency for founders, even the ones I work with today, to make big, lofty, noble promises.

但当情况变得艰难时，很容易就把这个旋钮调高了。

And then when times get a little tougher, it's very easy to turn that dial Yeah.

比如增加广告量，或者做其他类似的事情。

On the ad load or this or that.

你们希望如何在内部防止这种情况发生？

How do you hope to protect against that internally

是的

Yeah.

当你们应对不同的市场环境、竞争对手等时，如何做到这一点？

As you guys navigate different market conditions, competitors, etcetera?

因为当生意遇到困难时，这难道不是最容易按下按钮的做法吗？

Because is that not the easiest button to press when things get tough as a business?

所以，与完全依赖广告的公司相比，我们有趣的地方在于拥有多种不同的商业模式。

So, you know, the thing that's interesting about us compared to fully ad supported companies that we have many different business models.

对吧？

Right?

我们有消费者订阅的商业模式。

We have a consumer subscription business model.

我们也有广告。

We have ads.

我们还有企业客户购买我们的服务。

We have B2B companies buying our services.

所以你们将会拥有内置约翰尼的设备

That's So you'll have devices with Johnny in

没错。

Exactly.

正是如此。

The Exactly.

这意味着我们拥有多种收入来源，因此无需过度依赖其中任何一种。

And so that means we have many different revenue streams, and that allows us not to have to max out any single one of them.

所以我认为这非常有帮助。

So I think that's really helpful.

第二点，虽然听起来老生常谈，但我们从研究实验室创立之初就建立了一种非常强大的文化，始终致力于做对世界有益、与使命一致的事情。

The second thing, and that's going to sound cliche, is just like we have a very, very strong culture that was, you know, built from the very early days of the research lab to really do things that are for the good of the world and align with the mission.

所以，即使当我加入并开始考虑做广告时，我也并没有直接去写代码。

So, even when, you know, I arrived and we started thinking about doing ads, like I didn't just like go and started coding.

我实际上与公司里的大量人员进行了圆桌讨论。

I actually did roundtables with tons of people at the company.

我认为有100多人参加了不同的圆桌会议，因为大家都非常热衷于探讨如何以避免这些陷阱的方式来做广告。

I think we had 100 people show up in different roundtables because people were passionate about figuring out how we were going to do ads in a way that doesn't fall into those pitfalls.

这就是我们制定这些原则的由来。

And that's how we came up with the principles.

我们实际上和研究人员、以及非常关心此事的人一起共同构建了这些原则。

We actually like build them with researchers, with people at the table who cared enormously.

因此，我认为我们有一种能够抵御这种倾向的文化，如果我们在那个方向上走得太远，大家一定会强烈反对。

And so I think we have a culture that has the antibodies against that, and will absolutely revolt if like, you know, we go too far in that direction.

从根本上说，我们这样做是为了打造更好的产品，以便让每个人都能使用这款产品。

And fundamentally, like we're doing it to make a better product so that like, we can put that product in the hands of everyone.

如果我们在这条路上走得太远，那就根本不可能实现。

If we go too far towards that, that's just not gonna happen.

所以

So

是的

Yeah.

我认为这些原则至关重要，因为我在Snap经历的一件事是，当那些曾经作为抗体或保护者的人离开后，新人能依据什么来判断呢？

I think those principles are imperative because one thing I witnessed at Snap is that when people leave who may have been those antibodies or protectors, what can someone new point to?

是的。

Yeah.

一个新人，新的产品经理来了。

Someone who is new, a new PM arrives.

而产品经理通常都会想，我要让某些东西增长并看起来更好。

And as PMs do, they're like, I wanna make something go up and look good.

如果没有可以参考的依据，他们就会按自己的方式做事。

And if there isn't something to point to, then they're gonna do their job.

那么，这些原则具体有哪些呢？

And so what what are some of those principles?

这些原则是公开的吗？

Are those public?

是的。

Yeah.

广告相关的吗？

The ad ones?

这其实是故意为之的。

And that was actually very intentional.

我们把它们公开，以便大家都能看到并承担责任。

It was like we made them public so that we have excelled You're accountable.

能力。

Ability.

当然。

Absolutely.

其中一些原则是，广告不会影响大型语言模型的结果。

And so a couple of them are like, you know, ads are not going to influence the results of like the LLMs.

我认为这是最重要的一个。

That's, I would say, the most important one.

另一个是，广告始终会与内容明确区分开来。

Another one is that ads are always going to be very clearly separate and delineated from the content.

所以，你完全清楚自己能得到什么。

So, you know exactly what you're getting.

我们还有另一个关于数据控制的点。

We have another one that is all about data control.

我们当然不会把你的数据卖给广告商，而且我们还赋予你极大的控制权，包括完全删除所有用于广告的个人数据的能力。

So like we do not sell your data to advertisers, obviously, but we also give you a ton of control, including the ability to completely delete all of your personal data that's being used for ads.

所有这些控制措施都非常清晰、醒目，而且我们承诺永久坚持。

So all of those controls are like very clear, very prominent, and we've committed to them forever.

有趣的是，我认为谷歌也会说他们的广告是明确区分的。

So funny, I think Google would say that their ads are clearly delineated as well.

但它们只是越来越接近BlueLinx页面的结果，以至于你可能根本察觉不到有什么不对劲了。

But they just keep getting closer and closer to the results of the BlueLinx page to the point where you may not even know that something's bad anymore.

我的意思是，这不就是商业模式吗？

I mean, that's a business that's what?

已经有二十多年了。

Twenty something years old.

天啊。

Oh my god.

在亚马逊上，我觉得是的。

On Amazon, I feel like Yeah.

广告或赞助内容看起来和亚马逊推荐的一模一样。

The ad or sponsored looks exactly the same as Amazon's choice.

是的。

Yeah.

对。

Yeah.

所以我觉得这有点故意模棱两可。

And so it's kind of deliberately ambiguous, I think.

我们在原则中已经说得非常清楚了，你会在体验中看到的。

We made it pretty clear in our principles, and and you'll see the experience.

对吧？

Right?

我觉得这个体验非常、非常清晰。

Like, I think the experience is very, very clear.

我也认为

I also think that

嗯，我们不会因为付费而刻意追求。

Well, we won't seek as we pay.

但是，是的，这正是你的权利。

But, yes, That's to to your right.

我们还会在五年后的下一次面试中对你进行问责。

And we will also hold you accountable in our next interview in five years.

当然。

Absolutely.

随着那个小广告标识变得越来越小，我

As the little ad bug just gets smaller and I

我同意签署这个。

am I am fine signing up on that.

但你知道，另一件事是，有时它被视为一种必要的恶。

But you know, the other thing is like, is sometimes seen as like this necessary evil.

当我审视自己使用Instagram的情况时，比如，我从那些广告里买了好多东西。

When I look at my own usage of like, let's say Instagram, like, I buy so much stuff from those ads.

它让我发现了大量优质产品。

It has allowed me to discover a ton of good products.

它对那些 otherwise 无法被发现的中小企业产生了巨大的积极影响。

It has massive positive impact for SMBs who would not get discovered otherwise.

所以实际上我认为，如果我们做得好、有品位，并且在保护信任和隐私的前提下，这个产品可以带来非常积极的影响。

So I actually think that if we do it well, tastefully, in a way that preserves trust, in a way that preserves privacy, I think we can have a very positive impact with this product.

而且，把这一点免费交给成千上万的人使用，也是一个额外的好处。

And that's, you know, added benefit of putting that in the hands of tons of people for free.

也许这是一个更技术性的问题，但它关系到人们会如何感受这个产品。

Maybe a bit of a more technical question, but I think it speaks to how people will feel about the product.

你面前有两条发展方向。

You've kind of got two directions ahead of you.

你可以选择Facebook路线，也就是你参与打造的Facebook广告模式，或者选择Google广告路线。

You could go the Facebook, which you helped build the Facebook ads route, or you could go the Google ads route.

对大多数人来说，他们可能不太理解这两者的区别，但对Facebook来说，比如我会告诉Facebook：我想触达俄亥俄州喜欢Target的妈妈们。

And to most people, they probably don't understand the difference, but for Facebook, it's, you know, I tell Facebook, I wanna reach moms in Ohio who like Target.

这是基于你的经验吗？

Speaking from experience?

是的。

Yes.

这是我们做的播客。

That's our podcast.

这是你的副业。

That's your side hustle.

Facebook就会自己去执行，利用社交图谱等等一切。

And and Facebook just goes and does it, and they use the social graph and all that.

你很清楚，你参与打造了这个系统。

You know this, you help build this.

还有谷歌的做法，它是基于关键词、竞价机制的，不像Facebook那样与用户身份深度绑定，也不会主动帮你去推断目标。

There's also the Google approach, which is keyword based, auction based, and is not tied to identity in the same way, and is not like, we go figure it out for you.

是的。

Yeah.

你们打算选择哪个方向？

Which direction are you guys gonna go?

因为你的前Facebook同事Vijay正在帮你们组建广告团队，他可是来自Facebook的。

Because I mean, your ex Facebook, Vijay who's helping you build the ads team is ex Facebook.

我的意思是，OpenAI里有很多前Facebook的人。

I mean, there's a lot of ex Facebook people at at OpenAI.

所以，你认为你们会更倾向于谷歌模式，还是Facebook模式？

So are you gonna go the more Google route or the more Facebook route, do you think?

首先，我们拥有丰富的广告经验，但我们都来这里是为了构建一种与过去完全不同的广告模型，以适应我们当前的产品。

Well, first off, we have a lot of ads experience, but we all came here to build a very different ads models than what we built in the past that's appropriate for the product we have in front of us.

我认为，如果你看看ChatGPT，谷歌的模式会更相关。

And I think if you look at ChatGPT, the Google model is a lot more relevant.

当你在谈论某件事时，显然你有某种意图，而ChatGPT正在帮助你。

Like you're talking about something, you're clearly having some kind of intent, and like ChatGPT is helping you.

因此，你可以想象，与这些内容相关的广告会表现得更好，对用户来说也更相关。

And so you can imagine ads like being related to that content performing a lot better and being a lot more relevant for the user.

所以，我们肯定会从这里开始，随着时间推移，可能会出现一些意图驱动较弱、更开放的场景。

So, well, definitely starting there, You can imagine over time that there might be surfaces that are more like less intent driven and more open

去，暂停。

to, Pause.

是的，没错。

Yeah, exactly.

暂停，那种情况可能更接近Facebook的模式。

Pause where it might be, you know, closer to the Facebook model.

但我们目前非常专注于，与你所讨论内容达到最高级别的相关性。

But we're we're very much starting with, like, max level of relevance for the content that you're talking about.

这在科技界已经变成一个热门话题。

This has just become such a conversation in tech.

我的意思是，我刚去过达沃斯，那里每个人都在谈论广告和人工智能。

I mean, I was just at Davos, and it was like everyone was talking about ads and AI.

我甚至和你的一位前同事Boz聊了聊这件事，他说在聊天机器人场景下，谷歌的模式最有道理，但人们也确实会担心广告会以某种方式与LLM的输出挂钩。

And I was even catching up with Boz, one of your former colleagues about it and he was saying that that basically the Google model makes the most sense in a in a chatbot context, but also that, you know, people are rightfully gonna be still worried that the ad is somehow tied to the LLM's output.

无论你怎么设计，都是如此。

No matter what you do to design it.

谷歌的德米斯也说了类似的话。

And Demis from Google was kind of saying a similar thing.

我很想知道，即使你从后台说明了它的运作方式，你认为人们还是会认为，不，这实际上受到了我告诉ChatGPT的内容的影响吗？

And I'm curious, like, do you think you can even if you're saying in the background, this is how it works, do you think people will still there's gonna be a challenge with people still thinking, no, this is actually like influenced by what I'm telling ChatGPT?

我认为这确实是一个风险。

I think that's definitely a risk.

我的意思是，Facebook就是一个很好的例子。

I mean, you know, Facebook is a good example.

Facebook并没有把你的数据卖给广告商，但很多人就是这么认为的。

Facebook didn't sell your data to advertisers, but tons of people think they did.

Facebook并不会监听你的麦克风，但很多人却认为它会，这算是一种认知风险吗？

Facebook doesn't listen to your microphone and tons of people And think they so, is that a perception risk?

当然，我们正在通过教育手段在产品中尽一切努力来提升用户的认知。

For sure, we're doing everything we can in the products through education to educate people.

我认为更有帮助的是，因为我们拥有一个对话式产品，人们实际上会与模型互动，从而理解其中的原理。

The thing that I think is also helpful is because we have a conversational product, I think people will actually talk to the model and realize what's going on.

我给你举个例子。

So I'll give you an example.

因为模型并不了解广告，如果你问模型：嘿，我刚刚在屏幕上看到了一些广告。

Because the model is not aware of the ads, if you ask the model like, Hey, I just see some ads on my screen.

你能告诉我更多关于它们的信息吗？

Can you tell me more about them?

模型会告诉你：我其实看不到广告，因为我们没有提供这些信息。

The model is gonna tell you, I actually don't see the ads, because we're not surfacing that.

但如果你想让我看到广告，点击这个按钮即可。

But if you want me to see them, click this button.

这会让人们意识到，默认情况下，模型是看不到这些的。

And that reinforces to people like, oh, by default, the model is not seeing that.

所以这是一个很好的例子。

So that's like a good example.

此外，当模型说‘你可以查看广告并告诉我你的看法’时，它会做出完全中立的回应。

And then on top of that, once it says, okay, you can actually look at the ads and tell me what you think, the model is gonna respond something that is completely neutral.

这意味着在某些情况下，模型会说：‘我看到了这两家公司的广告。’

So that means that in some cases, the model is gonna say, well, I see the ads for these two companies.

但实际上

But actually

这个不错，但那个其实不是，我认为这会进一步强化中立性。

That one is good, but that one is actually not And really I think that's gonna reinforce neutrality.

我认为一些品牌会被这种做法吓到。

Now, I think that some brands are gonna get spooked by this.

这会让一些品牌生气。

That's gonna piss some brands off.

是的

Yeah.

但我认为，我们生活在一个透明和真实至关重要的世界里。

But but I also think that we live in a world where, like, transparency and authenticity is what counts.

因此，如果一个品牌声誉不好，导致Chad GPT说了些负面评价，它们应该去改善自身，而不是怪我们。

And so if a brand has a bad reputation so that Chad GPT ends up saying something bad, like, they should fix that instead of instead of blaming us.

所以我真的认为，与模型的这种对话将以其他任何方式都无法做到的方式强化中立性，你知道的。

So I actually think the discussion with the model will reinforce the neutrality in a way that nothing else could, you know.

有意思。

Interesting.

我认为，将广告植入如此私人的产品中，我确实相信广告可以帮助你发现更多可能感兴趣的内容。

I think the placement of the ads inside of a product that is so personal, I definitely am a believer that ads can be helpful and show you more of what you might interest be interested in.

毕竟，它们能解决你生活中的问题。

After all, do solve problems in your life.

我认为，人们对此如此担忧和顾虑的原因，几乎纯粹是出于原则。

I think I'm gonna argue that the reason people are so apprehensive and concerned about this is almost just on principle.

比如说，为什么 arguably 人类历史上最棒的产品

Like, why does arguably the best product in human history

你是说那是厨师的 e p t 吗？

Are you saying that's chef's e p t?

可能是消费级 AI。

Maybe consumer AI.

我来接手。

I'll take it.

是的。

Yeah.

消费级 AI。

Consumer AI.

好的。

Alright.

人类历史上最棒的产品之一。

One of the best products in human history.

为什么这需要广告？

Why does that need ads?

你知道的。

You know?

我认为Facebook在提供服务的成本上有所不同。

And I think there is a difference with Facebook with the cost of providing the service.

你知道，运行这些大量令牌的成本，与运行照片信息流或文字信息流之类的东西相比。

You know, the cost of burning all those tokens as opposed to running a a photo news feed or a text news feed or something like that.

但如果你没有广告，长期来看用户会更少吗？

But would you have fewer users long term if you didn't have ads?

既然争论的焦点是可及性，我觉得会有同样多的人使用它，因为它足够好、足够重要。

Given that the argument is about accessibility, I feel like the same number of people would use it just because of how good and how important it is.

这让我想起了苹果的事情。

Well, this brought me back to the Apple thing.

事实上，使用Facebook的人比苹果与ATT之争开始时还要多。

Like, turns out more people use Facebook than when the whole Apple war over ATT started.

对吧？

Right?

所以实际上人们确实会使用带有广告的免费服务。

So it's like people do actually use free services with ads.

是的。

Yeah.

你知道，广告虽然名声不好，但最终人们还是会喜欢使用有广告的服务。

Like, you know, ads have a bad rep, but ultimately people like end up, you know, enjoying services with ads.

我想说的是，你谈的是当前的世界，但想想人工智能发展有多快。

And the thing I would say is like, you're saying that in the current world, but think about how fast intelligence is progressing.

未来，免费版本中拥有最高水平的智能将变得越来越重要。

It's gonna become more and more like important to have the highest levels of intelligence in the free tier.

我认为，如果你不提供广告，最终会出现分化：付费用户将获得强大的智能支持，获得超能力。

And I think if you don't offer ads, what's gonna end up happening is you're gonna have a divergence where the people on paid tiers are gonna get tons of great intelligence are gonna be empowered with these superpowers.

而免费用户只能获得还不错的体验，但两者之间会有很大差距。

And then people into free tier are gonna have, like okay experience, but with a pretty big gap.

所以我们不想制造这种差距。

So that's the gap we don't wanna create.

我们希望

We want

提供一个简化版的库或互联网或

Having a stripped down version of the library or the internet or

是的。

Yeah.

或者就只是，是的。

And or just like Yeah.

你知道的，旧模型、较低的智能、较少的工具使用，所有这些。

You know, older models, lower intelligence, lower tool use, like all of these things.

所以，在我看来，是的，我们可能仍然有很多用户，但我希望这些用户能拥有尽可能高的智能，以便他们在生活中做更多事情。

And so, you know, to me, it's like, yeah, maybe we have, we still have a lot of users, but I want those users to have like as much intelligence as possible so they can do things in their life.

我认为，如果我们做得好的话，通过广告来支持这一点，将会是一种中性甚至有益的提升。

And I think, you know, having ads to support that will like be neutral to an enhancement if we do our job well.

我只是希望你们所有人，包括Anthropic，能更明确地说出这样做的商业原因。

I just kind of wish that all of you, Anthropic included, would just actually say like the business reasons for this too more.

我明白你们总在谈普及访问，但你们也得赚钱啊。

Like, I get the democratization access thing a lot, but, I mean, you guys gotta make money.

你们的大多数用户，其实是在给你们烧钱。

Most of your users, you are they cost you money.

对吧？

Right?

因为他们是

Because they're

免费的，哦，

free Oh,

我们为免费用户花的钱简直是个天文数字。

We we spend a fortune on Yeah.

所以这种局面迟早会发生。

Free so this was inevitable that this was gonna happen.

顺便说一下，Anthropic 那件事，我注意到他们说未来可能会做广告，而且他们那种说法是

And by the way, Anthropix thing, noticed they say we may do ads in the future and the thing where they're kind of

羞耻？我的意思是，Elise 说过，人们会给自己留一些余地。

shaming Yeah, mean, you Elise was saying, you know, people leave themselves some openness.

有趣的是，我们当初宣布广告时，是带着非常明确的原则的。

The funny thing the funny thing in this is like, we went out announcing ads with very strong principles.

他们却反过来批评广告，但又为广告留了很大的余地

They went out criticizing ads, but leaving the door very open for them

去做

to do

它在

it in the

未来。

future.

并且说他们要做电商和购物，但那篇博客里难道没有一个论点是：看，我们得为这个付费。

And saying they're gonna do commerce and shopping in But that same blog like, isn't there an argument to just saying like, look, we have to pay for this.

而且，我们支付这种方式就是通过做广告，总不能让数十亿人付费使用ChatGPT吧。

And like, this is the way we're gonna pay for this is we're gonna do ads in you're not gonna have billions of people paying for ChatGPT.

根本不存在这样的消费者，你在Facebook再清楚不过了，任何消费者产品如果没有广告，都不可能达到那样的规模。

There's just no consumer, you know this well at Facebook, like no consumer product reaches that scale without ads.

是的。

Yeah.

所以，不如坦然接受这一点，承认这就是现实。

And so just kind of embracing that and saying like, that's what it is.

而且，几年前萨姆还称这是最后的手段，但你们是一家庞大的公司，迟早要上市，必须赚钱。

And like, yeah, Sam called it a last resort a couple of years ago, but like, you all are a huge company and you're gonna probably IPO and you need to make money.

这其实就是关键，对吧？

And like, that's kind of the thing, right?

是的，但我认为这并不是为了赚钱而赚钱。

Yeah, but I think it's not like making money for the sake of making money.

我认为赚钱是为了能够继续以这种智能水平服务数十亿用户。

I think it's making money so that we can continue serving billions of users with that level of intelligence.

我们也可以通过在免费版中降低智能水平来赚钱。

We could also make money by just not serving that much intelligence in the free tier.

我更愿意通过这种方式赚钱并扩大访问权限，而不是通过在免费版中削减智能来赚钱。

And like, I'd rather, you know, I'd rather make money this way and expanding access than making money by just, you know, saving on intelligence in the free tier.

我想退一步说。

I wanna go back one step.

哪些免费用户给你带来了最大的成本？

Which free users are costing you the most money?

他们在做什么？

What are they doing?

嗯，我们能看到所有的使用场景。

Well, you know, I mean, we see all use cases.

对吧？

Right?

比如图像生成，还有大量创建图片。

Like, image gen and, like, creating tons of images.

这太贵了。

That's that's expensive.

我们看到人们使用大量工具、个性化功能，比如你使用了多少内存、存储了多少数据。

We see people, like, you know, using lots of tools, personalization, like, you know, how much you use memory and how much you have stored in your memory.

所有这些因素都会影响我们的成本。

Like all of these things matter to, you know, how much it costs us.

而且，所有这些功能都是对产品的绝佳补充。

And again, all of these things are great addition to the product.

比如，你使用带记忆和不带记忆的ChatGPT，会发现巨大的差异，对吧？

Like if you use ChatGPT with and without memory, you see a massive difference, right?

所以我们希望让更多人使用这些功能，即使Pulse的运营成本高得惊人。

So these are things that we want to expand access to even Pulse, you know, Pulse costs a fortune to deliver.

我们会在你没有主动要求的情况下，在后台工作，每天早上为你提供高度个性化的服务。

It's like, we do work in the background without you asking us anything to deliver something hyper personalized to you every morning.

这需要花费大量资金。

That costs a lot of money.

目前，它仅对Pro用户开放，我们希望将来也能让Plus用户使用，并最终向免费用户开放。

Right now, it's available to Pro, we want to make it available to Plus and at some point, we want to make it available to Free as well.

这些服务成本很高，但我们希望让每个人都能享受到。

Like, these are the kinds of services that are costly, but we wanna bring to everyone.

Pulse让我非常感兴趣，因为我长期报道Facebook，马克多年来一直谈论的个性化报纸概念，现在似乎终于在Pulse上实现了。

Pulse is so interesting to me because having covered Facebook for so many years, that whole idea of the personal newspaper that Mark was talking about for all those years, it feels like it's finally coming to life with Pulse.

是的。

Yeah.

我特别喜欢的一点是，当我还在Facebook时，我们试图为你提供个性化报纸，但当时受限于世界上已有的内容。

And and the thing I love about it is that, you know, when I was at Facebook and we're trying to give you the personalized newspaper, like, we were limited by the content that was created in the world.

对吧？

Right?

而Pulse不同，我们会自己生成内容，这才是真正神奇的地方。

With Pulse, we create the content, And that's what's super magical.

前几天，它给我推荐了一篇关于时尚与奇幻阅读交叉点的内容，这种主题太小众了，根本不会有现成的文章，但它却推断出我对这两者都有兴趣。

Like, the other day, it served me something about the intersection of fashion and fantasy reading, which, like, there's no article that's gonna like, it's too niche, but it figured out she has these two interests.

我们将撰写一篇关于奇幻书籍中时尚仪式的完整文章。

We're gonna create an entire article about fashion ritual in fantasy books.

我当时就想，天啊，世界上可能只有我在意这个。

And I was like, oh my god, I'm the only person in the world that cares.

这太棒了。

And this is amazing.

所以，斐济是一种我们尚未察觉的特殊类型的极客。

So And Fiji is a very special type of nerd that we have not yet detected.

确实如此。

Indeed.

所以，我看了之后心想，以前这完全是不可能的。

And so, you know, I looked at that and I was like, that would have been totally impossible before.

我觉得，这个工具真正懂我。

And like, I'm like, this tool sees me.

是的。

Yeah.

所以我认为这就是Pulse的魔力。

And so I I think that's the magic of Pulse.

而且还有更多即将推出，因为目前我们只是触及了皮毛。

And there's so much more coming because right now we're just scratching the surface.

我非常好奇今年ChatGPT作为产品会走向何方。

I'm so curious about where ChatGPT is going as a product this year.

它会如何演变呢？

What's what how's it gonna evolve?

是的。

Yes.

巨大的转变是从一个仍然本质上是被动的聊天机器人，演变为一个个人超级助手，帮助你完成生活中的所有事务。

So the big transformation is going from what is still essentially a reactive chatbot to a personal super assistant that allows you to get get life done, like all the tasks in your life done.

这意味着大量的过渡。

And that means a lot of transition.

这意味着要从一个被动的工具转变为一个主动的工具。

That means like going from reactive tool to a proactive tool.

我们刚刚谈到了Pulse，那就是主动性。

We are talking about Pulse a second ago, that's the proactivity.

这意味着从一个本质上基于文本的工具，转变为更加视觉化和多媒体化的工具。

It means going from a tool that's self fundamentally text based to being much more visual and multimedia.

这意味着从你主要获取信息的地方，转变为你可以设置任务并让它自动完成的地方。

It means going from a place where you're mostly just getting information to a place where you actually set up a task and it just gets it done.

因此，我们现在在产品中看到了许多这样的变化，但还需要进一步加强。

And so there's a bunch of these things that we are now seeing coming in the product, but need to really amplify.

我认为，到今年年底，它应该真正成为你的助手，让你以更高效的方式完成生活的方方面面。

And I think, you know, by the end of the year, it should really feel like your assistant that allows you to get every part of your life done in a much more efficient way.

当你思考‘你的助手’这个概念时，我觉得人们对它可能有不同的理念。

As you think about the idea of your assistant, I feel like there are different philosophies around what that might mean.

比如，我是一个总是想学习新事物、保持高度好奇心的人。

Like, I'm someone who's always trying to learn new things and and be very curious.

有时候我发现，当我打开Reels或TikTok时，它给我推荐的都是太多相似的内容。

And sometimes I find when I open Reels or TikTok that it's showing me too much of the same.

是的。

Yep.

你知道的。

You know?

当你思考个人智能在向你展示更多你喜爱和感兴趣的内容时，你是怎么看待等待的呢？这些内容可能从参与度角度看风险更高，但作为一个人，看到并深入探索它们或许能促进你的成长？

How do you think about waiting in terms of the personal intelligence showing you more of what you like and what you're interested in and more of things that are probably riskier from an engagement point of view, but might be better as a person to have seen and potentially dig into to grow as a person?

是的，这是个非常好的问题。

Yeah, that's such a great question.

你知道，在Pulse中，我们几乎每天都会问你：嘿，我们给你展示了这五个主题。

You know, in Pulse, we actually ask you like pretty much every day like, hey, we showed you these five themes.

你希望看到更多内容吗？

Is there more that you would like to see?

我们甚至会向你展示一些你可能想拓展的相关主题。

We even show you like, these are some related themes that you might want to expand into.

你怎么看？

What do you think?

而且，我们拥有一个你可以引导的对话式产品，这与所有其他产品相比是一个巨大的差异。

And so the fact again that we have a conversational product that you can steer is like a massive difference compared to all of these other products.

因为在其他产品中，提供反馈的方式非常有限。

Because in these other products, like there's very few like ways to give feedback.

无非就是点赞或叉掉。

It's like a thumbs up, an X out.

但你可以直接说：嘿，我看了很多关于奇幻阅读的内容，能不能给我展示一些别的东西？

But you can say, hey, like, you know, I've seen a lot about fantasy reading, show me a little bit more about something else.

而在我们的产品中，你可以这么做。

And in our products, you can.

所以我们正在大力投入可引导性。

So we're leaning really hard into steerability.

同时也在探索界面设计，让我们能逐步拓展你的视野，向你展示一些相关联的内容，让你告诉我们你是否喜欢。

And also like figuring out interfaces where we can like start to expand your world a little bit and show you things that are adjacent, so that you can tell us if you like it or not.

这对科技来说是一个巨大的转变。

This is a big shift for tech.

对吧？

Right?

我的意思是，作为一个长期关注Facebook和Instagram的人，而且一直听亚当和莎拉等人讲话，我总觉得，可以说这是由Facebook主导的一种观念：你点击什么，你就想要什么，而这对你来说才是更好的。

I mean, as someone who's been following Facebook and Instagram forever, and, you know, listening to people like Adam and Sarah speak for a long time, I think there's often been this idea, I would say led by Facebook, that what you click is what you want, and that's more of what's good for you.

而我现在非常惊讶地看到，Instagram竟然让你自己选择算法中显示的内容，这可能会导致你使用产品的时长减少。

And now I was very surprised to see and and delighted that Instagram's letting you choose what what appears in the algorithm, which could ostensibly result in you using the product less.

你认为这种变化为什么现在会发生？

Why why do you think that change is happening now?

更广泛地说，随着时间推移。

More broadly too, just across time.

是的。

Yeah.

对。

Yeah.

你知道，我认为这是因为人们意识到，某一天你一旦陷入这些信息茧房，就很难再脱身了。

You know, I think it's happening because people have realized that, like, at some point you go into these rabbit holes and and it's very hard to get out of that.

我认为技术已经发展到让我们能够进行引导，而这是以前不可能做到的。

And I think the technology has evolved that allows us to steer, which wasn't possible before.

而且我认为你提到的‘使用产品的时间会减少’这一点，其实没人真的相信。

And I also think that your point on like, in my results in using the product less, like I actually think no one believes that.

我认为每个人都相信，如果我们允许你进行引导并拓展你的视野，你实际上会发现一些让产品在长期来看更具吸引力的内容。

I think everyone believes that if we allow you to steer and allow you to expand your horizons, you're actually going to discover things that makes product like, know, much more engaging over time.

因为重复的东西迟早会让人感到无聊。

Because repetition at some point gets boring.

如果你不知道如何摆脱这种状态，最终你会减少使用产品。

And if you don't know how to get out of it, you know, you end up using the product less in that case.

所以我认为这实际上是一种帮助人们更好地表达需求的方式，但我也认为这对平台有利，最终来说。

So I think actually it is a way to get people to like sell their needs better, but also I think it it does accrue to the platform, so ultimately

你们所讨论的，似乎与传统的以广告驱动的商业模式本质上是相悖的。

And what you guys are talking about, it seems inherently at odds with the traditional ads driven business model.

对吧？

Right?

我知道你们说我们不会以在ChatGPT中的使用时长为目标。

And I know you guys say we will not goal for time spent in ChatGPT.

是的。

Yeah.

但随着广告规模扩大，你们有信心这仍然能成立吗？

But as ads scale, do you have confidence that that can actually still be true?

你们能同时做到你们所说的——实现可控性，最终打造一个让人感觉更好、更高效、不断成长的产品，同时还能维持广告驱动的商业模式吗？

That you can do what you're talking about and have steerability and ultimately try to make a product that makes people feel better and more productive and and grow and still have an ad driven business model?

我只是觉得，在科技史上，这还从未发生过。

I just don't think that's has that happened yet in the history of tech?

有趣的是，如果看看我们的收入预测（我不会详细透露），广告在我们的收入结构中将长期保持少数地位。

Well, so the thing that's interesting is if you look at our revenue projection, which I'm not gonna give you in details, ads will remain a minority of our revenue in our projections and

你们说的是几年，还是多久？

For like years or how long are we talking?

很多很多年。

For many, many years.

好的。

Okay.

而且我认为可能是永远如此。

And I I think probably forever.

所以，你知道，这真的吗？

And so, you know, it is Really?

你并不认为它会成为主营业务的多数，对吧？

You don't see it being the majority of the business No.

在某个时候吗？

At some point?

不会。

No.

我认为它是一个非常有益的商业模式，正如我们提到的原因。

I think that it's a very helpful business model for the reasons we mentioned.

但我真的不认为它会成为主营业务的多数。

But I actually really don't think it's going to be the majority of the business.

当你观察B2B市场，看看那些公司为借助AI加速自身发展而付费的情况时。

When you look at the market on the B2B side for, you know, companies paying to accelerate themselves with AI.

即使考虑到那些希望获得最大程度智能支持的人群的消费倾向，我认为这仍然只是少数现象。

When you look even at like propensity to spend for some people who are going to want the maximum access of intelligence, like I actually think it stays a minority.

这非常有帮助，因为它让我们能够专注于打造真正顶尖的产品。

And that's very helpful because that allows us to stay focused on like really just building the absolute best product.

最终，我认为即使在那些公司选择优化使用时长等方向的情况下，我也怀疑这是否真的对他们长期有利，甚至在收入方面也是如此。

And ultimately, you know, I think even in the cases where companies have gone the way of like optimizing for time spent, etcetera, I'm not sure that serves them well for the long run, even in terms of revenue.

因此，我想我们已经从这些经验中学到了很多。

And so, you know, I think we've learned from all of these lessons.

我认为，专注于打造有用的产品，无论如何都是实现长期收入最大化的最佳途径。

And I think optimizing for having a product that's useful is likely the best path to optimizing for revenue long term anyways.

而且我认为，我们都清楚，广告是一个绝佳的商业模式，甚至可以说是历史上最优秀的商业模式之一——通过拍卖机制，广告主愿意支付恰好足够的费用来触达特定的个人。

And I think, you know, we all know ads are a great business, if not one of the best business models ever created with the auction where advertisers are gonna pay exactly as much as they need to to reach this exact person.

但我记得萨姆曾明确表示，广告将是最后的手段。

But I I remember Sam went so far as to say that ads would be a last resort.

那里发生了什么变化？

What changed there?

主要的变化是我们之前讨论过的，那就是作为免费层级中最优秀的智能服务，成本非常高，而我们意识到广告与我们的使命是一致的。

The main thing is what we've we've discussed, which is like, it costs a lot to serve as the best intelligence in the free tier and the realization that ads were mission aligned.

我知道，创始人很容易从这样的角度出发，觉得广告是一种必要的恶。

You know, I think I think it's easy for founders to start from that place, like, you know, oh, ads is kind of a necessary evil.

我认为他逐渐意识到，我们其实有一种非常好的方式，通过广告将更多智能服务提供给尽可能多的人。

And I think he realized over time that, you know, we really had like a very good way of serving more intelligence to as many people as possible with ads.

所以这是亚历克斯的。

So it's Alex

如果没有广告，他就不会找到这件夹克。

wouldn't have found this jacket if it wasn't for ads.

那并不是

That's not

真的。

true.

我这么说，看起来很棒。

I'm so it looks great.

看起来太棒了。

It looks fantastic.

我Actually买了那件。

I actually bought that one.

没错。

That's right.

就像我们协调了一样。

It's like we coordinated.

对。

Yeah.

是的。

Yeah.

所以你的头衔是应用首席执行官。

So your title is CEO of applications.